Production of charged pions, kaons, and (anti-)protons in Pb-Pb and inelastic pp collisions at √sNN = 5.02 TeV

Midrapidity production of π ± , K ± , and ( p ) p measured by the ALICE experiment at the CERN Large Hadron Collider, in Pb-Pb and inelastic pp collisions at √ s NN = 5 . 02 TeV, is presented. The invariant yields are measured over a wide transverse momentum ( p T ) range from hundreds of MeV / c up to 20 GeV / c . The results in Pb-Pb collisions are presented as a function of the collision centrality, in the range 0–90%. The comparison of the p T -integrated particle ratios, i.e., proton-to-pion ( p /π ) and kaon-to-pion ( K /π ) ratios, with similar measurements in Pb-Pb collisions at √ s NN = 2 . 76 TeV show no signiﬁcant energy dependence. Blast-wave ﬁts of the p T spectra indicate that in the most central collisions radial ﬂow is slightly larger at 5.02 TeV with respect to 2.76 TeV. Particle ratios ( p /π , K /π ) as a function of p T show pronounced maxima at p T ≈ 3GeV / c in central Pb-Pb collisions. At high p T , particle ratios at 5.02 TeV are similar to those measured in pp collisions at the same energy and in Pb-Pb collisions at √ s NN = 2 . 76 TeV. Using the pp reference spectra measured at the same collision energy of 5.02 TeV, the nuclear modiﬁcation factors for the different particle species are derived. Within uncertainties, the nuclear modiﬁcation factor is particle species independent for high p T and compatible with measurements at √ s NN = 2 . 76 TeV. The results are compared to state-of-the-art model calculations, which are found to describe the observed trends satisfactorily.


I. INTRODUCTION
Previous observations at the Relativistic Heavy-Ion Collider (RHIC) and at the CERN Large Hadron Collider (LHC) demonstrated that in high-energy heavy-ion (A-A) collisions, a strongly interacting quark-gluon plasma (sQGP) [1][2][3][4][5] is formed. It behaves as a strongly coupled near-perfect liquid with a small viscosity-to-entropy ratio η/s [6]. The experimental results have led to the development and adoption of a standard theoretical framework for describing the bulk properties of the QGP in these collisions [7]. In this paradigm, the beam energy dependence is mainly encoded in the initial energy density (temperature) of the QGP. After formation, the QGP expands hydrodynamically as a near perfect liquid before it undergoes a chemical freeze-out. The chemical freezeout temperature is nearly beam-energy independent for centerof-mass energy per nucleon pair larger than 10 GeV [7,8].
The hadronic system continues to interact (elastically) until kinetic freeze-out. We report in this paper a comprehensive study of bulk particle production at the highest beam energy for A-A collisions available at the LHC. We probe the highest QGP temperature, to further study this paradigm and address its open questions.
Transverse momentum distributions of identified particles in Pb-Pb collisions provide information on the transverse expansion of the QGP and the freeze-out properties of the ensuing hadronic phase. By analyzing the p T -integrated yields in Pb-Pb collisions it has been shown that hadron yields in high-energy nuclear interactions can be described assuming their production at thermal and chemical equilibrium [9][10][11][12], with a single chemical freeze-out temperature, T ch ≈ 156 MeV, close to the one predicted by lattice QCD calculations for the QGP-hadronic phase transition, T c = (154 ± 9) MeV [13]. Indeed, the Pb-Pb data from LHC Run 1 [14] showed an excellent agreement with the statistical hadronization model with the exception of the proton and antiproton, (K * )K * and multistrange particle yields [9,12]. The deviation could be in part due to interactions in the hadronic phase, which result in baryon-antibaryon annihilation that is most significant for (anti-)protons [15][16][17][18]. Proposed explanations for the observed discrepancy with respect to the thermal model predictions can be found in Refs. [18][19][20][21][22]. Moreover, at √ s NN = 2.76 TeV the proton-to-pion [(p + p)/(π + + π − ) ≡ p/π ] ratio exhibits a slight decrease with centrality and a slightly lower value than measured at RHIC. New measurements at √ s NN = 5.02 TeV, which exploit the currently highest medium density, could provide an improved understanding of the particle production mechanisms [22]. The spectral shapes at low p T (p T <2 GeV/c) in central Pb-Pb collisions at √ s NN = 2.76 TeV showed a stronger radial flow than that measured at RHIC energies, in agreement with the expectation based on hydrodynamic models [14,23]. The results for identified particle production at low p T and higher √ s NN are useful to further test hydrodynamic predictions.
At intermediate p T (2 − 10 GeV/c), the particle ratios experimentally show the largest variation and in particular for the baryon-to-meson enhancement several new hadronization mechanisms have been proposed [24][25][26]. In the most central Pb-Pb collisions at √ s NN = 2.76 TeV, the p/π ratio reaches values larger than 0.8 for p T ≈ 3 GeV/c, which surpass those for inelastic pp collisions at the same energy [27,28]. An intermediate p T enhancement of heavier hadrons over lighter hadrons is expected from the collective hydrodynamic expansion of the system alone [29][30][31]. In coalescence models [32][33][34], which requires radial flow as well, baryon-to-meson ratios are further enhanced at intermediate p T by the coalescence of lower p T quarks that leads to a production of baryons (3 quarks) with larger p T than for mesons (2 quarks). The baryon-to-meson ratio decreases at high p T and reaches the values observed in pp collisions as a consequence of the increasing importance of parton fragmentation. The observation of a qualitatively similar enhancement of the kaon-to-pion [(K + + K − )/(π + + π − ) ≡ K/π ] ratio in central Pb-Pb collisions with respect to inelastic pp collisions [28,35] supports an interpretation based on the collective radial expansion of the system that affects heavier particles more.
For high p T (p T >10 GeV/c), measurements of the production of identified particles in Pb-Pb collisions relative to inelastic pp collisions contribute to the study of hard probes propagating through the medium. This offers the possibility to determine the properties of the QGP like the transport coefficientq [36] and the space-time profile of the bulk medium in terms of local temperature and fluid velocity [37]. The modification of particle production is quantified with the nuclear modification factor, R AA , defined as where d 2 N AA /(dyd p T ) is the particle yield in nucleus-nucleus collisions and σ pp is the production cross section in pp collisions. The average nuclear overlap function is represented by T AA and is obtained from a Glauber model calculation [38]. It is related to the average number of binary nucleon-nucleon collisions N coll , and the total inelastic nucleon-nucleon cross section, σ NN INEL = (67.6 ± 0.6) mb at √ s NN = 5.02 TeV [39], as T AA = N coll /σ NN INEL . Several measurements of R AA at high p T for different √ s NN [40][41][42][43][44][45][46] support the formation of a dense partonic medium in heavy-ion collisions where hard partons lose energy via a combination of elastic and inelastic collisions with the constituents of the QGP [47]. Results from Pb-Pb collisions at √ s NN = 2.76 TeV showed that within uncertainties, the suppression is the same for pions, kaons and (anti-)protons [28]. Moreover, the inclusive charged-particle nuclear modification factor measured in Pb-Pb collisions at 5.02 TeV shows that the suppression continues to diminish for p T above 100 GeV/c [48] while the suppression of jets saturates at a value of 0.5 [49]. Particle production at high transverse momentum has also been studied as a function of the Bjorken energy density [50] and path length [51][52][53].
The results show interesting scaling properties which can be further tested using LHC data at higher energies.
In this paper, the measurement of p T spectra of π ± , K ± and (p)p in inelastic pp and Pb-Pb collisions at √ s NN = 5.02 TeV over a wide p T range, from 100 MeV/c   for pions, 200 MeV/c for kaons, and 300 MeV/c for (anti-)protons, up to 20 GeV/c for all species, are presented. Particles are identified by combining several particle identification (PID) techniques based on specific ionization energy loss (dE/dx) and time-of-flight measurements, Cherenkov radiation detection and the identification of the weak decays of charged kaons via their kink-topology. The article is organized as follows: Sec. II outlines the analysis details including the track and event selections as well as the particle identification strategies. The obtained results are discussed in Sec. III. Section IV presents the comparison of data with model predictions. Finally, Sec. V contains a summary of the main results.

II. DATA ANALYSIS
In this paper the measurements obtained with the central barrel of the ALICE detector, which has full azimuthal coverage around midrapidity, |η| < 0.8 [54], are presented. A detailed description of the ALICE detector can be found in Ref. [55].
The pp results were obtained from the analysis of ≈ 1.2 × 10 8 minimum bias pp collisions, collected in 2015. The Pb-Pb analysis with ITS and TOF uses ≈ 5 × 10 6 minimum bias Pb-Pb collisions, collected in 2015. The Pb-Pb analysis where PID is provided by the TPC, the high momentum particle identification (HMPID) detector and the kink decay topology requires more statistics and uses the full data sample collected in 2015 corresponding to ≈ 6.5 × 10 7 Pb-Pb collisions.
Both in pp and Pb-Pb collisions, the interaction trigger is provided by a pair of forward scintillator hodoscopes, the V0 detectors, which cover the pseudorapidity ranges 2.8 < η < 5.1 (V0A) and −3.7 < η < −1.7 (V0C) [56]. The minimum bias trigger is defined as a coincidence between the V0A and the V0C trigger signals. The V0 detector signals, which are proportional to the charged-particle multiplicities, are used to divide the Pb-Pb event sample into centrality classes, defined in terms of percentiles of the hadronic cross section [38]. A Glauber Monte Carlo model is fitted to the V0 amplitude distribution to compute the fraction of the hadronic cross section corresponding to any given range of V0 amplitudes. The 90-100% centrality class has substantial contributions from QED processes (≈ 20%) [38] and its low track multiplicity presents difficulties in the extraction of the trigger inefficiency; it is therefore not included in the results presented here. Also, an offline event selection is used to remove beam background events. It employs the information from two zero degree calorimeters (ZDCs) positioned at 112.5 m on either side of the nominal interaction point. Beam background events are removed by using the V0 timing information and the correlation between the sum and the difference of times measured in each of the ZDCs [55].
The central barrel detectors are located inside a solenoidal magnet providing a magnetic field of 0.5 T and are used for tracking and particle identification. The innermost barrel detector is the inner tracking system (ITS) [57], which consists of six layers of silicon devices grouped in three detector systems (from the innermost outwards): the silicon pixel detector (SPD), the silicon drift detector (SDD), and the silicon strip detector (SSD). The time projection chamber (TPC), the main central-barrel tracking device, follows outwards. The results are presented for primary particles, defined as particles with a mean proper lifetime τ > 1 cm/c which are either produced directly in the interaction or from decays of particles with τ < 1 cm/c, restricted to decay chains leading to the interaction [58]. To limit the contamination due to secondary particles and tracks with wrongly associated hits and to ensure high tracking efficiency, tracks are required to cross at least 70 TPC readout rows with a χ 2 normalized to the number of TPC space-points ("clusters"), χ 2 /NDF, lower than 2. Tracks are also required to have at least two hits reconstructed in the ITS out of which at least one is in the SPD layers and to have a distance of closest approach (DCA) to the interaction vertex in the direction parallel to the beam axis (z), |DCA z | < 2 cm. A p T -dependent selection on the DCA in the transverse plane (DCA xy ) of the selected tracks to the primary vertex is also applied [59]. Furthermore, the tracks associated with the decay products of weakly decaying kaons ("kinks") are rejected. The latter selection is not applied in the study of kaon production from kink decay topology. The primary vertex position is determined from tracks, including short track segments reconstructed in the SPD [60]. The position of the primary vertex along the beam axis is required to be within 10 cm from the nominal interaction point. The position along z of the SPD and track vertices are required to be compatible within 0.5 cm. This ensures a uniform acceptance and reconstruction efficiency in the pseudorapidity region |η| < 0.8 and rejects pileup events in pp collisions. Different PID detectors are used for the identification of the different particle species. Ordering by p T , from lowest to highest, the results are obtained using the dE/dx measured in the ITS and the TPC [61], the time of flight measured in the time-offlight (TOF) detector [62], the Cherenkov angle measured in the high-momentum particle identification detector (HMPID) [63] and the TPC dE/dx in the relativistic rise region of the Bethe-Bloch curve. The performance of these devices is reported in Ref. [55].

A. Particle identification strategy
For the analysis presented here, pions, kaons, and (anti-)protons have been identified following the same analysis techniques as in the previous ALICE measurements. The ITS, TPC (low p T ) and TOF analyses are described in Refs. [14,64,65], while the HMPID and TPC (high p T ) analyses are documented in Refs. [28,35,66]. The kink analysis is described in Ref. [59]. In this paper, only the most relevant aspects of each specific analysis are described.
In most analyses, the yield is extracted from the numberof-sigma (N σ ) distribution. This quantity is defined as where i refers to a given particle species (i = π , K, p), signal is the detector PID signal (e.g., dE/dx), and signal i and σ i are the expected average PID signals in a specific detector and its standard deviation, respectively. Figure 1 shows the pion-kaon and kaon-proton separation power as a function of p T for ITS, TPC, TOF, and HMPID. The separation power is defined as follows: Note that the response for the individual detectors is momentum (p) dependent. However, since results are reported in transverse momentum bins, the separation power as a function of p T has been evaluated, averaging the momentumdependent response over the pseudorapidity range |η| < 0.5. In Table I  a. ITS analysis. The four outer layers of the ITS provide specific energy-loss measurements. The dynamic range of the analog readout of the detector is large enough [67] to provide dE/dx measurements for highly ionizing particles. Therefore, the ITS can be used as a standalone low-p T PID detector in the nonrelativistic region where the dE/dx is proportional to 1/β 2 . For each track, the energy loss fluctuation effects are reduced by using a truncated mean: the average of the lowest two dE/dx values in case four values are measured, or a weighted sum of the lowest (weight 1) and the second lowest (weight 1/2), in case only three values are available.
The plane (p; dE/dx) is divided into identification regions where each point is assigned a unique particle identity. The identity of a track is assigned based on which dE/dx curve the track is closest to, removing in this way the sensitivity to the dE/dx resolution. To reject electrons, a selection on |N π σ | < 2, is applied. Using this strategy, it is possible to identify π and K with an efficiency of about 96-97% above p T = 0.3 GeV/c, and (p)p with an efficiency of 91-95% in the entire p T range of interest. In the lowest p T bin, the PID efficiency reaches ≈ 60%, ≈ 80%, and ≈ 91% for pions, kaons, and (anti-)protons, respectively. By means of this technique it is possible to identify π ± , K ± , and (p)p in Pb-Pb (pp) collisions in the p T ranges 0.1-0.7 GeV/c, 0.2-0.5 (0.6) GeV/c, and 0.3-0.6 (0.65) GeV/c, respectively.
b. TOF analysis. The analysis with the TOF detector uses the subsample of tracks for which a time measurement with TOF is available. The time of flight t TOF is the difference between the measured particle arrival time τ TOF and the event time t 0 , namely t TOF = τ TOF − t 0 . In the ALICE experiment, the t 0 value can be obtained with different techniques [68]. The best precision on the t 0 evaluation is obtained by using the TOF detector itself. In this case, the t 0 is obtained on an event-by-event basis by using a combinatorial algorithm that compares the measured τ TOF with the expected one under different mass hypotheses. The procedure to evaluate t 0 with the TOF detector is fully efficient if enough reconstructed tracks are available, which is the case of the 0-80% Pb-Pb collisions. The resolution on the t 0 evaluated with the TOF detector is better than 20 ps if more than 50 tracks are used for its determination. This improvement with respect to Run 1 performance [68] is due to improved calibration procedures carried out during Run 2. Overall the TOF signal resolution is about 60 ps in central Pb-Pb collisions. In pp and 80-90% Pb-Pb collisions the measurement of the event time relies on the T0 detector (σ t T0 ev ≈ 50 ps) [68] or, in case it is not available, on the bunch crossing time, which has the worst resolution (≈ 200 ps). The PID procedure is based on a statistical unfolding of the time-of-flight N σ distribution. For each p T bin, the expected shapes for π ± , K ± , and (p)p are fitted to the t TOF distributions, allowing the three particles to be distinguished when the separation is as low as ≈ 2σ . An additional template is needed to account for the tracks that are wrongly associated with a hit in the TOF. The templates are built from data as described in Ref. [14]. For this purpose  the length of measured tracks is used to compute a realistic distribution of the expected time of arrival for each mass hypothesis and the signal shape is reproduced by sampling the parametrized TOF response function (described by a Gaussian with an exponential tail) obtained from data. Since the rapidity of a track depends on the particle mass, the fit is repeated for each mass hypothesis. TOF analysis makes identification of π ± , K ± , and (p)p in Pb-Pb (pp) collisions possible in the p T ranges 0.60-3.50 GeV/c, 1.00 (0.65)−3.50 GeV/c and 0.80-4.50 GeV/c, respectively. c. TPC analysis. The TPC provides information for particle identification over a wide momentum range via the specific energy loss [55]. Up to 159 space-points per trajectory can be measured. A truncated mean, utilizing 60% of the available clusters, is employed in the dE/dx determination [61]. The dE/dx resolution for the Minimum Ionizing Particle (MIP) is ≈ 5.5% in peripheral and ≈ 6.5% in central Pb-Pb collisions. Particle identification on a track-by-track basis is possible in the region of momentum where particles are well separated by more than 3σ . This allows the identification of pions, kaons and (anti-)protons within the transverse momentum ranges 0.25-0.70 GeV/c, 0.25-0.45 GeV/c, and 0.45-0.90 GeV/c, respectively.
The TPC dE/dx signal in the relativistic rise region (3 < βγ 1000), where the average energy loss increases as ln(βγ ), allows identification of charged pions, kaons, and (anti-)protons from p T ≈ 2-3 GeV/c up to p T = 20 GeV/c. The first step of the TPC high-p T analysis is the calibration of the PID signal; a detailed description of the the dE/dx calibration procedure can be found in Refs. [28,35]. Particle identification requires precise knowledge of the dE/dx response and resolution σ . This is achieved using the PID signals of pure samples of secondary pions and protons originating from K 0 S and decays as well as a sample of tracks selected with TOF. In addition, measured K 0 S spectra are used to further constrain the TPC charged kaon response [28]. For different momentum intervals, a sum of four Gaussian functions associated with the pion, kaon, proton and electron signals is fitted to the dE/dx distribution.
d. HMPID analysis. The HMPID performs identification of charged hadrons based on the measurement of the emission angle of Cherenkov radiation. Starting from the association of a track to the MIP cluster centroid one has to reconstruct the photon emission angle. Background, due to other tracks, secondaries and electronic noise, is discriminated exploiting the Hough Transform Method (HTM) [69]. Particle identification with the HMPID is based on statistical unfolding. In pp collisions, a negligible background allows for the extraction of the particle yields from a three-Gaussian fit to the Cherenkov angle distributions in a narrow transverse momentum range. In the case of Pb-Pb collisions, the Cherenkov angle distribution for a narrow transverse momentum bin is described by the sum of three Gaussian distributions for π ± , K ± , and (p)p for the signal and a sixth-order polynomial function for the background [28]. This background is due to misidentification in the high occupancy events: the larger the angle, the larger the probability to find background clusters arising from other tracks or photons in the same event. This background is uniformly distributed on the chamber plane. The resolution in Pb-Pb events is the same as in pp collisions (≈ 4 mrad at β ≈ 1). In this analysis, the HMPID provides results in pp and Pb-Pb collisions in the transverse momentum ranges 1.5-4.0 GeV/c for π ± and K ± , and in 1.5-6.0 GeV/c for (p)p.
e. Kink analysis. In addition to the particle identification techniques mentioned above, charged kaons can also be identified in the TPC using the kink topology of their two-body decay mode (e.g., K → μ + ν μ ) [59]. With the available statistics, this technique extends PID of charged kaons up to 4 GeV/c in pp collisions and up to 6 GeV/c in Pb-Pb collisions. The kink analysis reported here is applied for the first time to Pb-Pb data. For the reconstruction of kaon kink decays, the algorithm is implemented within the fiducial volume of the TPC detector (130 < R < 200 cm), to ensure that an adequate number of clusters is found to reconstruct the tracks of both the mother and the daughter with the necessary precision to be able to identify the particles. The mother tracks of the kinks are selected using similar criteria as for other primary tracks, except that the minimum number of TPC clusters required are 30 instead of 70, because they are shorter compared to the primary ones. Assuming the neutrino to be massless, the invariant mass of the decayed particle (M μν ) is estimated from the charged decay product track and the momentum of the neutrino as reported in Ref. [59]. The main background is from charged pion decays, π → μ + ν μ (B.R. = 99.99%), which also gives rise to a kink topology. A proper q T selection, where q T is the transverse momentum of the daughter track with respect to the mother's direction at the kink, can separate most of the pion kink background from the kaon kinks. Since the upper limit of q T values for the decay channels π → μ + ν μ and K → μ + ν μ are 30 MeV/c and 236 MeV/c, respectively, a selection of q T > 120 MeV/c rejects more than 80% (85% in pp collisions) of the pion background. For further removal of the contamination from pion decays, an additional selection on kink opening angle, as reported in Ref. [59], has been implemented. Finally, the TPC dE/dx of the mother tracks is required to have |N K σ | < 3, which improves the purity of the sample. After these selections, the purity ranges from 99% at low p T to 92% (96% in pp collisions) at high p T according to Monte Carlo studies. The remaining very low background is coming from random associations of charged tracks reconstructed as fake kinks. After applying all these topological selection criteria, the invariant mass of kaons (M μν ) obtained from the reconstruction of their decay products integrated over the above mentioned mother momentum ranges for pp and Pb-Pb collisions are shown in Fig. 2.

B. Correction of raw spectra
To obtain the p T distributions of primary π ± , K ± , and (p)p, the raw spectra are corrected for PID efficiency, misidentification probability, acceptance, and tracking efficiencies, following the procedures described in Ref. [14] for the ITS, TPC (low p T ) and TOF, in Ref. [28] for the HMPID and TPC (high p T ) and in Ref. [59] for the kink analysis. The acceptance, reconstruction, and tracking efficiencies are obtained from Monte Carlo simulated events generated with PYTHIA 8.1 (Monash 2013 tune) [70] for pp collisions and with HIJING [71] for Pb-Pb collisions. The particles are propagated through the detector using the GEANT 3 transport code [72], where the detector geometry and response, as well as the data taking conditions, are reproduced in detail. Since GEANT 3 does not describe well the interaction of low-momentum p and K − with the material, a correction to the efficiencies is estimated using GEANT 4 and FLUKA, respectively, which are known to describe such processes better [14,[73][74][75]. The PID efficiency and the misidentification probability are evaluated by performing the analysis on the Monte Carlo simulation, which requires that the simulated data are first tuned to reproduce the real PID response for each PID technique. The contamination due to weak decays of light flavor hadrons (mainly K 0 S affecting π ± spectra, and + affecting (p)p spectra) and interactions with the material has to be computed and subtracted from the raw spectra. Since strangeness production is underestimated in the event generators and the interactions of low p T particles with the material are not properly modeled in the transport codes, the secondary-particle contribution is evaluated with a data-driven approach. For each PID technique and species, the contribution of feed-down in a given p T interval is extracted by fitting the measured distributions of DCA xy of the tracks identified as the given hadron species. The DCA xy distributions are modeled with three contributions: primary particles, secondary particles from weak decays of strange hadrons and secondary particles produced in the interactions with the detector material. Their shapes are extracted for each p T interval and particle species from the Monte Carlo simulation described above. The contribution of secondaries and after (lower) the topological selection. The peak centered at M μν = 0.49 GeV/c 2 is for the decay channel K → μ + ν μ (B.R. = 63.55%), whereas the peak centered at M μν = 0.43 GeV/c 2 is for the decay channel K → π + π 0 (B.R. = 20.66%), whose invariant mass is calculated with the wrong mass hypothesis.
is different for each PID analysis due to the different track and PID selections and is more important at low p T . The measured Pb-Pb spectra are then normalized to the number of events in each centrality class.
The spectra measured in pp collisions are also normalized to the number of inelastic collisions obtained from the number of analyzed minimum bias events corrected with an inelastic normalization factor of 0.757 (± 2.51%), defined as the ratio between the V0 visible cross section and the inelastic pp cross section at √ s = 5.02 TeV [39].

C. Systematic uncertainties
The evaluation of systematic uncertainties follows the procedures described in Ref. [14] for the ITS, TPC (low p T ), and TOF analyses, in Ref. [28] for the HMPID and TPC (high p T ) analyses and in Ref. [59] for the kink analysis. The main sources of systematic uncertainties, for each analysis, are summarized in Tables II and III, for the Pb-Pb and pp analyses, respectively. Sources of systematic effects such as the different PID techniques, the feed-down correction, the imperfect description of the material budget in the Monte Carlo simulation, the knowledge of the hadronic interaction cross section in the detector material, the TPC-TOF and ITS-TPC matching efficiency, and the track selection have been taken into account. The systematic uncertainties related to track selection were evaluated by varying the criteria used to select single tracks (number of reconstructed crossed rows in the TPC, number of available clusters in the ITS, DCA xy and DCA z , χ 2 /NDF of the reconstructed track). The ratio of the corrected spectra with modified selection criteria to the default case is computed to estimate the systematic uncer-tainty for a given source. A similar approach is used for the evaluation of the systematic uncertainties related to the PID procedure. The uncertainties due to the imperfect description of the material budget in the Monte Carlo simulation is estimated varying the material budget in the simulation by ±7%.
To account for the effect related to the imperfect knowledge of the hadronic interaction cross section in the detector material, different transport codes (GEANT3, GEANT4, and FLUKA) are compared. Finally, the uncertainties due to the feed-down correction procedure are estimated for all analyses by varying the range of the DCA xy fit, by using different track selections, by applying different cuts on the (longitudinal) DCA z , and by varying the particle composition of the Monte Carlo templates used in the fit.
For the ITS analysis, the standard N σ method is compared with the yields obtained with a Bayesian PID technique [76]. Moreover, the Lorentz force causes shifts of the cluster position in the ITS, pushing the charge in opposite directions when switching the polarity of the magnetic field of the experiment (E × B effect) [14]. This effect is not fully reproduced in the Monte Carlo simulation and has been estimated by analyzing data samples collected with different magnetic field polarities. To estimate possible systematic effects deriving from signal extraction in the low p T TPC analysis, the yield was computed by varying the selection based on the number of TPC crossed rows from 70 to 90 and the yield was computed from the sum of the bin content of the N σ distribution in the range [−3, 3], instead of fitting.
The systematic uncertainty was obtained from the comparison to the nominal yield. Regarding the TPC analysis at high p T , the imprecise knowledge of both the Bethe-Bloch TABLE II. Main sources and values of the relative systematic uncertainties (expressed in %) of the p T -differential yields of π ± , K ± , and (p)p obtained in the analysis of Pb-Pb collisions. When two values are reported, these correspond to the lowest and highest p T bin of the corresponding analysis, respectively. If only one value is reported, then the systematic uncertainty is not p T -dependent. If not specified, then the uncertainty is not centrality-dependent. The first three systematic uncertainties are common to all PID techniques. The maximum (among centrality classes) total systematic uncertainties and the centrality-independent ones are also shown.   TABLE III. Main sources and values of the relative systematic uncertainties (expressed in %) of the p T -differential yields of π ± , K ± , and (p)p obtained in the analysis of pp collisions. When two values are reported, these correspond to the lowest and highest p T bin of the corresponding analysis, respectively. If only one value is reported, then the systematic uncertainty is not p T -dependent. The first three systematic uncertainties are common to all PID techniques. In the last row, the total systematic uncertainty is reported.  a TOF time response function with varied parameters. For the HMPID analysis, the selection on the distance between the extrapolated track point at the HMPID chamber planes and the corresponding MIP cluster centroid, d MIP−trk , is varied by ±1 cm to check its systematic effect on the matching of tracks with HMPID signals. Moreover, the systematic bias due to the background fitting, which represents the largest source, is estimated by changing the fitting function: from a sixth-order polynomial to a power law of the tangent of the Cherenkov angle. This function is derived from geometrical considerations [77]. For the kink analysis, the systematic uncertainties are estimated by comparing the standard spectra with the ones obtained by varying the selection on decay product transverse momentum, the minimum number of TPC clusters, kink radius and TPC N σ values of the mother tracks. By using the same methods as for the spectra, the systematic uncertainties for the p T -dependent particle ratios were computed to take into account the correlated sources of uncertainty (mainly due to PID and tracking efficiency). Finally, for both p T -dependent spectra and ratios the particlemultiplicity-dependent systematic uncertainties, those that are uncorrelated across different centrality bins, were determined.
The improved reconstruction and track selection in the analysis of pp and Pb-Pb data at √ s NN = 5.02 TeV lead to reduced systematic uncertainties as compared to previously published results at √ s NN = 2.76 TeV.

III. RESULTS AND DISCUSSION
The measured p T spectra of π ± , K ± , and (p)p from the independent analyses have to be combined in the overlapping ranges using a weighted average with the systematic and statistical uncertainties as weights. All the systematic uncertainties are considered to be uncorrelated across the different PID techniques apart from those related to the ITS-TPC matching efficiency and the event selection. The correlated systematic uncertainties have been added in quadrature after the spectra have been combined. For a given hadron species, the spectra of particles and antiparticles are found to be compatible, and therefore all spectra reported in this section are shown for summed charges. Figure 3 shows the combined p T spectra of π ± , K ± , and (p)p measured in 0-90% Pb-Pb and inelastic pp collisions at √ s NN = 5.02 TeV. Results for Pb-Pb collisions are presented for different centrality classes. Scaling is applied in the plots to improve spectra visibility. In the low p T region, the maximum of the spectra is pushed toward higher momenta while going from peripheral to central Pb-Pb events. This effect is mass dependent and can be interpreted as a signature of radial flow [14]. For high p T , the spectra follow a power-law shape, as expected from perturbative QCD (pQCD) calculations [78].
The p T -integrated yields, dN/dy, and the average transverse momentum, p T , are determined for the different centrality classes using an extrapolation to p T = 0. The extrapolation procedure is performed after fitting the measured spectra with Boltzmann-Gibbs Blast-Wave [79] (for Pb-Pb) or the Lévy-Tsallis [80,81] (for pp) functions. In the most central Pb-Pb collisions (0-5%), the extrapolated fractions of the total yields are 5.84%, 5.20%, and 3.72%, for pions, kaons, and (anti-)protons, respectively. The fractions increase as centrality decreases, reaching 8.63%, 9.36%, and 10.73% in the most peripheral collisions (80-90%). In pp collisions the fractions are 8.59%, 9.98%, and 12.61% for pions, kaons, and (anti-)protons, respectively. The systematic uncertainties are then propagated to the p T -integrated yields and mean transverse momentum. For the uncertainty on dN/dy, the fit is performed with all data points shifted up by their full systematic uncertainties. To estimate the uncertainty on p T , points in the 0-3 GeV/c range are shifted up and down within their systematic uncertainty to obtain the softest and hardest spectra. The maximum difference (in absolute value) between the integrated quantities obtained with the standard and modified spectra are included as part of the systematic uncertainty.
Additionally, different functions 1 were used to perform the extrapolation and the largest differences were added to the previous contributions.
The statistical uncertainties on the dN/dy and p T values are evaluated propagating the uncertainties on the fit parameters obtained directly from the fit procedure. The procedure described above is repeated using the systematic uncertainties uncorrelated across different centrality bins to extract the centrality uncorrelated part of the systematic uncertainties on the p T -integrated particle yields and the average transverse momenta.
In Table IV, the dN/dy and p T are shown for Pb-Pb and pp collisions, respectively. For Pb-Pb collisions the values are given for different centrality ranges.

A. Particle production at low transverse momentum
The Boltzmann-Gibbs blast-wave function is a threeparameter simplified hydrodynamic model in which particle production is given by [79] 1 Lévy-Tsallis (Pb-Pb only); Boltzmann-Gibbs blast-wave (pp only); m T -exponential: Ax × exp(− √ x 2 + m 2 /T ), where A is a normalization constant, T the temperature, and m the mass; Fermi-Dirac 044907-9 The velocity profile ρ is given by where β T is the radial expansion velocity, m T the transverse mass (m T = m 2 + p T 2 ), and T kin the temperature at the kinetic freeze-out, I 0 and K 1 are the modified Bessel functions, r is the radial distance in the transverse plane, R is the radius of the fireball, β s is the transverse expansion velocity at the surface, and n is the exponent of the velocity profile.
To quantify the centrality dependence of spectral shapes at low p T , the Boltzmann-Gibbs blast-wave function has been simultaneously fitted to the charged pion, kaon and (anti-)proton p T spectra, using a common set of parameters but different normalization factors and masses. Although the absolute values of the parameters have a strong dependence on the p T range used for the fit [14], the evolution of the parameters with √ s NN can still be compared across different collision energies by using the same fitting ranges. The present analysis uses the same p T intervals employed for fitting as in a previous publication [14], namely, 0.5-1 GeV/c, 0.2-1.5 GeV/c, and 0.3-3 GeV/c for charged pions, kaons, and (anti-)protons, respectively. Figure 4 shows the ratios of the spectra to results of the fits for different centrality classes and particle species. If the shape of the p T distributions over the full measured p T range was purely driven by the collective radial expansion of the system, then the functions determined by fitting the data in a limited p T range would be expected to describe the spectral shapes in the full p T range. Within uncertainties, this is only observed for the proton p T spectra (up to 4 GeV/c) in 0-20% Pb-Pb collisions. A different situation is observed for pions where, due to their small mass and the large centralitydependent feed-down contribution from resonance decays, the agreement with the model is worse than that observed for kaons and (anti-)protons.
The p T interval where the model describes the data within uncertainties gets wider going from peripheral to central Pb-Pb collisions.
FIG. 5. Average expansion velocity ( β T ) and kinetic freeze-out temperature (T kin ) progression from the simultaneous Boltzmann-Gibbs blast-wave fit to π ± , K ± , and p(p) spectra measured in Pb-Pb collisions at √ s NN = 5.02 and 2.76 TeV [14]. The correlated uncertainties from the global fit are shown as ellipses. The elliptic contours correspond to 1σ uncertainties, with statistical and systematic uncertainties being added in quadrature.
In Table V the blast-wave fit parameters β T , T kin and n in Pb-Pb collisions at √ s NN = 5.02 TeV, for different centrality classes, are listed. Figure 5 shows the correlation between β T and T kin , both obtained from the simultaneous fit for Pb-Pb collisions at √ s NN = 2.76 TeV and 5.02 TeV. For Pb-Pb collisions at √ s NN = 5.02 TeV, β T increases with centrality, reaching β T = 0.663 ± 0.003 in 0−5% central collisions, while T kin decreases from T kin = (0.161 ± 0.006) GeV to T kin = (0.090 ± 0.003) GeV, similarly to what was observed at lower energies. This can be interpreted as a possible indication of a more rapid expansion with increasing centrality [4,14]. In peripheral collisions this is consistent with the expectation of a shorter lived fireball with stronger radial pressure gradients [82]. The value of the exponent of the velocity profile of the expansion, n, is about 0.74 in central collisions and it increases up to 2.52 in peripheral collisions (see Table V). The values of n in peripheral collisions increase with respect to those in central collisions to reproduce the TABLE V. Results of the combined Boltzmann-Gibbs blast-wave fits to the particle spectra measured in Pb-Pb collisions at √ s NN = 5.02 TeV, in the p T ranges 0.5-1 GeV/c, 0.2-1.5 GeV/c, and 0.3-3.0 GeV/c for π ± , K ± , and (p)p, respectively. Values in parenthesis refer to the ratios to the values in Pb-Pb collisions at √ s NN = 2.76 TeV [14]. The charged particle multiplicity values are taken from Refs. [84,85]. power-law tail of the p T spectra. Finally, in the most central Pb-Pb (0-5%) collisions the difference of the average transverse velocity between the two collision energies is ≈ 2.4 standard deviations.
The value at 5.02 TeV is ≈ 2% larger than that measured at 2.76 TeV, whereas the kinetic freeze-out temperature results are slightly smaller at larger collision energy but the difference is not significative. Just for the most peripheral collisions the kinetic freeze-out temperature is slightly higher at 5.02 TeV than that at 2.76 TeV. This is in contrast with our interpretation for central collisions where a larger volume has the kinetic freeze-out later allowing the kinetic temperature to decrease further. It is worth questioning whether the blast wave formalism is applicable also for these smaller system and it will be interesting to see if models, which can also describe small systems, can explain this changing pattern. Moreover, we note that event and geometry biases may also play a role in the peripheral Pb-Pb collisions [83]. Figure 6 shows the p T for charged pions, kaons, and (anti-)protons as a function of the charged particle multiplicity density dN ch /dη at midrapidity in Pb-Pb collisions at √ s NN = 5.02 and 2.76 TeV. Going from inelastic pp collisions to peripheral and central Pb-Pb collisions, the p T increases with dN ch /dη . The rise of the average p T gets steeper with increasing hadron mass, this effect is consistent with the presence of radial flow. Within uncertainties and for comparable charged particle multiplicity densities, the results for both energies are consistent for 20-90% Pb-Pb collisions. For 0-20% Pb-Pb collisions, p T is slightly higher at 5.02 TeV than at 2.76 TeV. The increase originates from the low p T part of the spectra. Again, this is an effect consistent with a stronger radial flow in Pb-Pb collisions at the highest collision energy. Figure 7 shows the p T -integrated particle ratios, K/π and p/π , as a function of dN ch /dη in Pb-Pb at on the integrated ratios have been evaluated using the uncertainties on the p T -dependent ratios, taking into account the part of the uncertainties correlated among the different particle species. No significant energy dependence is observed, indicating that there is small or no dependence of the hadrochemistry on the collision energy. The K/π ratio hints at a small increase with centrality. The effect is consistent with the observed increase of strange to nonstrange hadron production in heavy-ion collisions compared to inelastic pp collisions [86]. The p/π ratio suggests a small decrease with centrality. Using the centrality uncorrelated uncertainties, the difference between the ratio in the most central (0-5%) and peripheral (80-90%) collisions is ≈ 4.7 standard deviations, thus the difference is significant. The decreasing ratio is therefore consistent with the hypothesis of antibaryon-baryon annihilation in the hadronic phase [16][17][18][19]87,88]. The effect is expected to be less important for the more dilute system created in peripheral collisions.
Recently, a new procedure has been implemented to quantitatively estimate properties of the quark-gluon plasma created in ultrarelativistic heavy-ion collisions utilizing Bayesian statistics and a multiparameter model-to-data comparison [89].
The study is performed using a recently developed parametric initial condition model, reduced thickness event-byevent nuclear topology (T R ENTo) [90], which interpolates among a general class of energy-momentum distributions in the initial condition, and a modern hybrid model which The average transverse momentum as a function of dN ch /dη is quite well reproduced by the model. The model predicts that the kaon-to-pion ratio should decrease with increasing charged particle multiplicity density while data show an increase with dN ch /dη . Within uncertainties, the model agrees with the data for the most central Pb-Pb collisions. The trend of the proton-to-pion ratio is qualitatively well captured by the model but the values of the centrality-dependent ratios are overestimated. Figure 8 shows the K/π and p/π ratios as a function of p T for Pb-Pb collisions at √ s NN = 2.76 and 5.02 TeV. The results are also compared with inelastic pp collisions at √ s = 5.02 TeV. Within uncertainties, in the K/π ratio, no significant energy dependence is observed in heavy-ion data over the full p T interval. The ratios measured in 60-80% Pb-Pb collisions at both √ s NN agree within systematic uncertainties with that for inelastic pp collisions over the full p T range. Given that in pp collisions at LHC energies the ratio as a function of p T does not change with √ s [66], and given the similarity between pp and peripheral Pb-Pb collisions, the large difference observed is likely a systematic effect of the measurement and not a physics effect.

B. Intermediate transverse momentum
In general, the particle ratios exhibit a steep increase with p T going from 0 to 3 GeV/c while for p T larger than 10 GeV/c little or no p T dependence is observed. Going from peripheral to the most central Pb-Pb collisions, the ratios in the region around p T ≈ 3 GeV/c are continuously growing. A hint of an enhancement with respect to inelastic pp collisions is observed at p T ≈ 3 GeV/c. As pointed out in previous publications [14,28], the effect could be a consequence of radial flow which affects kaons more than pions.
The p/π ratios measured in heavy-ion collisions exhibit a pronounced enhancement with respect to inelastic pp collisions, reaching a value of about 0.8 at p T = 3 GeV/c. This is reminiscent of the increase in the baryon-to-meson ratio observed at RHIC in the intermediate p T region [45,91]. Such an increase with p T is due to the mass ordering induced by the radial flow (heavier particles are boosted to higher p T by the collective motion) and it is an intrinsic feature of hydrodynamical models. It should be noted that this is also suggestive of the interplay of the hydrodynamic expansion of the system with the recombination picture as discussed in the introduction. However, since recombination mainly affects baryon-to-meson ratios, it would not explain the bump which is also observed in the kaon-to-pion ratio. The shift of the peak towards higher p T in the proton-to-pion ratio is consistent with the larger radial flow measured in Pb-Pb at √ s NN = 5.02 TeV compared to the one measured at √ s NN = 2.76 TeV. The mass dependence of the radial flow explains also the observation that the maximum of the p/π ratio is located at a larger p T as compared to the K/π ratio. The radial flow is expected to be stronger in the most central collisions, this explains the slight shift in the location of the maximum when central and peripheral data are compared. Finally, particle ratios at high p T in Pb-Pb collisions at both energies become similar to those in pp collisions, suggesting that vacuumlike fragmentation processes dominate there [35]. For p T < 10 GeV/c, protons appear to be less suppressed than kaons and pions, consistent with the particle ratios shown in Fig. 8. The large difference between the suppression of different species is consistent with a mass ordering related to the radial flow. It is worth noting that 2.76 TeV measurements [92] showed that the mesons, including φ(1020), have smaller R AA than protons, indicating a baryon-meson ordering, so while there is a strong radial flow component, there are other effects affecting R AA in this p T region. At larger p T , all particle species are equally suppressed. Despite the strong energy loss observed in the most central heavy-ion collisions, particle composition and ratios at high p T are similar to those in vacuum. This suggests that jet quenching does not affect particle composition significantly.

C. Particle production at high transverse momentum
In the identified particle R AA for peripheral Pb-Pb collisions an apparent presence of jet quenching is observed (R AA < 1), although for similar particle densities in smaller systems (like p-Pb collisions) no jet quenching signatures have been reported [93]. It has been argued that peripheral A-A collisions can be significantly affected by event selection and geometry biases [83], leading to an apparent suppression for R AA even if jet quenching and shadowing are absent. The presence of biases on the R AA measurement in peripheral Pb-Pb collisions has been confirmed in Ref. [94]: the geometry bias sets in at mid-central collisions, reaching about 15% for the 70-80% Pb-Pb collisions. The additional effect of the selection bias becomes noticeable above the 60% percentile and is significant above the 80% percentile, where it is larger than 20%. All hard probes should be similarly affected [83], in particular, the leading pions, kaons and (anti-)protons reported in the present paper. Figure 10 shows

044907-14
No significant dependence on the collision energy is observed, as also been observed for unidentified charged particles [95].

IV. COMPARISON TO MODELS
The results for identified particle production have been compared with the latest hydrodynamic model calculations based on the widely accepted "standard" picture of heavy-ion collisions [96]. These models all have similar ingredients: an initial state model provides the starting point for a viscous hydrodynamic calculation, chemical freeze-out occurs on a constant temperature hyper-surface, where local particle production is modeled with a statistical thermal model, and finally, the hadronic system is allowed to reinteract. The models used are: iEBE-VISHNU hybrid model [29,30], McGill [31], and EPOS [97]. In the following, specific features of each of them are described: (i) The iEBE-VISHNU model is an event-by-event version of the VISHNU hybrid model [98], which combines (2 + 1) − d viscous hydrodynamics VISH2+1 [99,100] to describe the expansion of the sQGP fireball with a hadron cascade model (UrQMD) [101,102] to simulate the evolution of the system in the hadronic phase. The prediction of iEBE-VISHNU using either T R ENTo (Sec. III A) or a multiphase transport model (AMPT) [103] as initial conditions gives a good description of flow measurements in √ s NN = 2.76 TeV Pb-Pb collisions. T R ENTo parametrizes the initial entropy density via the reduced thickness function; AMPT constructs the initial energy density profiles using the energy decomposition of individual partons. Predictions by the iEBE-VISHNU hybrid model is available for p T up to 3 GeV/c. (ii) The McGill model initial conditions rely on a new formulation of the IP-Glasma model [104], which provides realistic event-by-event fluctuations and nonzero pre-equilibrium flow at the early stage of heavy-ion collisions. Individual collision systems are evolved using relativistic hydrodynamics with nonzero shear and bulk viscosities [105]. As the density of the system drops, fluid cells are converted into hadrons and further propagated microscopically using a hadronic cascade model [101,102]. The McGill predictions are available for p T up to 4 GeV/c and centralities 0-60%. (iii) The EPOS model in the version EPOS3 is a phenomenological parton-based model that aims at modeling the full p T range. EPOS is based on the theory of the Gribov-Regge multiple scattering, perturbative QCD, and string fragmentation [105]. However, dense regions in the system created in the collisions, the so-called core, is treated as a QGP and modeled with a hydrodynamic evolution followed by statistical hadronization. EPOS3 implements saturation in the initial state as predicted by the Color Glass Condensate model [106], a full viscous hydrodynamic simulation of the core, and a hadronic cascade, not present in the previous version of the model. EPOS3 implements also a new physics process that accounts for hydrodynamically expanding bulk matter, jets, and the interaction between the two, important for particle production at intermediate p T [107] and reminiscent of the recombination mechanism [32,33]. Figure 11 shows the ratios of the p T spectra in Pb-Pb collisions at √ s NN = 5.02 TeV to the models described above for p T < 4 GeV/c. In the low p T regime, one expects bulk particle production to dominate, so the absence of hard physics processes in the iEBE-VISHNU-T R ENTo, iEBE-VISHNU-AMPT, and McGill calculations is a minor issue. One observes that all models, in general, describe the spectra and the centrality dependence around p T ≈ 1 GeV/c within 20%. For p T < 3 GeV/c the agreement with data is within 30%. The models agree with the proton (kaon) data over a broader p T range than for kaons (pions). This mass hierarchy is expected from the hydrodynamic expansion, which introduces a mass dependence via the flow velocity -the larger the mass the larger the p T boost. Similarly, it can be noticed that for the most central collisions the models describe the data over a broader p T range than in peripheral ones. This is as expected from simple considerations. In central collisions, the system is larger and so the hydrodynamic expansion lasts longer, resulting in a stronger flow. At the same time, the fraction of the system involved in this expansion, the so-called core (e.g., the fraction of participant partons experiencing two or more binary collisions), is larger for the most central collisions.
One can conclude that all four model calculations qualitatively describe the centrality dependence of radial flow and how it is imprinted on the different particle species. Like the simplified blast-wave fits in Fig. 4, the two iEBE-VISHNU calculations also have difficulties to describe the very low p T (p T < 0.5 GeV/c) pion spectra. Figure 12 shows the ratios of the p T spectra in Pb-Pb collisions at √ s NN = 5.02 TeV to the EPOS3 model up to 10 GeV/c in p T . EPOS3 includes both soft and hard physics processes, which should give a better description of data at high p T and in peripheral collisions. However, its agreement with data is not significantly better than for the other models in the same p T interval (p T < 3 GeV/c) and at high p T , it is about a factor 2 off with respect to data. For completeness, Figs. 13, 14, and 15 show the comparison of the models with the p T dependent particle ratios. The larger proton-to-pion ratio in EPOS3 than observed in the data can be understood as due to the underestimated pion yield in the model (see Fig. 12).
To compare the energy evolution of the spectra between data and model, in Fig. 16 is shown the ratio of the π ± , K ± , and (p)p p T spectra measured at √ s NN = 5.02 TeV to those measured at √ s NN = 2.76 TeV, compared to the same ratios obtained from model predictions. For the McGill model, predictions at √ s NN = 2.76 TeV are currently not available. For central collisions, the agreement of the energy evolution in data and predictions is very good for both VISHNU initial-state models, while for peripheral collisions the AMPT initial conditions are better. For EPOS3 instead, a good agreement with data can be observed for both central and peripheral collisions. The comparison of model predictions to the ALICE measurements of anisotropic flow [108][109][110] can be useful to obtain tighter constraints on them.

V. CONCLUSIONS
In this paper, a comprehensive measurement of π ± , K ± and (p)p production in inelastic pp and 0-90% central Pb-Pb collisions at √ s NN = 5.02 TeV at the LHC is presented. A clear evolution of the spectra with centrality is observed, with a power-law-like behavior at high p T and a flattening of the spectra at low p T , confirming previous results obtained in Pb-Pb collisions at √ s NN = 2.76 TeV. These features are compatible with the development of a strong collective flow with centrality, which dominates the spectral shapes up to relatively high p T in central collisions. The p T -integrated particle ratios as a function of dN ch /dη in Pb-Pb at on the collision energy. A blast-wave analysis of the p T spectra gives an average transverse expansion velocity of β T = 0.663 ± 0.004 in the most central (0-5%) Pb-Pb collisions that is ≈ 2% larger than at √ s NN = 2.76 TeV, with a difference of ≈ 2.4 standard deviations between the two energies. The p T -dependent particle ratios (p/π , K/π ) show distinctive peaks at p T ≈ 3 GeV/c in central Pb-Pb collisions, more pronounced for the proton-to-pion ratio. Such an increase with p T is due to the mass ordering induced by the radial flow that would affect heavier particles more than lighter ones. The p T of the peak position increases slightly with energy, in particular for the proton-to-pion ratio, indicating that the initially hotter system is longer lived so that radial flow is stronger. At high p T , both particle ratios at √ s NN = 5.02 TeV are similar to those measured at √ s NN = 2.76 TeV and in pp collisions, suggesting that vacuumlike fragmentation processes dominate there. No significant evolution of nuclear modification at high-p T with the center-of-mass energy is observed.
At high p T , pions, kaons, and (anti-)protons are equally suppressed as observed at √ s NN = 2.76 TeV. This suggests that the large energy loss leading to the suppression is not associated with strong mass ordering or large fragmentation differences between baryons and mesons. Transverse momentum spectra and particle ratios in Pb-Pb collisions are compared to different model calculations based on the standard QGP picture, which are found to describe the observed trends satisfactorily. For p T < 3 GeV/c, all models agree with the data within 30%, at p T ≈ 1 GeV/c they describe the spectra and the centrality dependence within 20%.

ACKNOWLEDGMENTS
The ALICE Collaboration thanks all its engineers and technicians for their invaluable contributions to the construc-tion of the experiment and the CERN accelerator teams for the outstanding performance of the LHC complex. The ALICE Collaboration gratefully acknowledges the resources and support provided by all Grid centres and the Worldwide LHC Computing Grid (WLCG) collaboration. The ALICE Collaboration acknowledges the following funding agencies for their support in building and running the ALICE detec-